Getting Started

The platform can run entirely locally, and its isolated incident demo does not require an external LLM key.

Fastest path

python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
cp .env.example .env
uvicorn app.main:app --reload

Open:

• Web Console: http://127.0.0.1:8000/
• API docs: http://127.0.0.1:8000/docs

Docker

cp .env.example .env
docker compose up --build

The Compose setup includes a readiness health check and persistent data volume.

Offline Multi-Agent demo

Run:

python scripts/demo_incident.py

The scenario asks why recent import jobs failed and executes the real Data, Knowledge, Ops and Supervisor code paths using isolated data and a deterministic Fake LLM.

Use JSON output when you want machine-readable results:

python scripts/demo_incident.py --json

Model runtime

Natural-language features use an OpenAI-compatible endpoint:

LLM_BASE_URL=https://api.openai.com/v1
LLM_API_KEY=
LLM_MODEL=
EMBEDDING_MODEL=

Local compatible endpoints can leave LLM_API_KEY empty.

Authentication

For a shared demo environment:

AUTH_ENABLED=true
AUTH_TOKENS={"user-token":"alice:user","operator-token":"operator:operator","approver-token":"reviewer:approver","admin-token":"admin:admin"}

CONSOLE_USERNAME=demo
CONSOLE_PASSWORD=demo
CONSOLE_ROLE=admin
SESSION_MAX_AGE_SECONDS=43200

The Web Console uses an HttpOnly session cookie. API clients may still use Bearer tokens.

Next

• Architecture
• Demo Guide
• Interview Guide